Modern medical service systems are lacking in terms of offering real-time patient monitoring, diagnosis, and early detection of disease symptoms and problems in patient’s health state.
The objective of this paper was to design a model of adaptive hybrid clinical decision and prediction support system under a ubiquitous environment for well-being life care, particularly related to high-risk disease and metabolic syndrome with noninvasive data such as vital signal data, food nutrient data, and activity data using a neural network.
The model is designed using the complementary two-phase reverse neural network for being covered measure the problems of standard MLP model, which are local optimum, difficulty in modifying small sample size data, and one-direction learning. We determined that most of the simulation cases were satisfied by the two-phase reverse prediction neural network. In particular, small sample size of times series were more accurate than the standard MLP model.
The research suggests the best prescription for prevention of diseases related to metabolic syndrome and high risk disease in the U-hospital, home healthcare system, PERS (Personal Emergency Response System), and silver town healthcare for elder people and patients.